A media network built from acquired news sites says it attracts roughly 50 million pageviews a month. The number would make it larger than many newsrooms with recognizable brands, experienced reporters and substantial payrolls.
The twist is that a months-long Futurism investigation found Brown Brothers Media publishing AI-generated articles under invented bylines across dozens of sites. The reporters identified more than 50 fake writer personas and described a “robot team” whose staff were told to generate articles with AI and edit them to sound human-written. Futurism also reported that the company began deleting material after questions arrived.
The investigation is worth reading in full. The more useful question for content teams is not whether AI was involved. It is what separates accountable AI-assisted publishing from a system built on fabricated identities, borrowed domain authority and volume.
What the investigation found
Futurism reports that Brown Brothers Media acquired established publications, kept the authority associated with those domains and filled them with high volumes of AI-generated material. Articles appeared under fictional authors, while some staff worked inside a production process explicitly organized around automated generation.
The company says its network receives about 50 million pageviews each month, a self-reported figure that has not been independently audited in the available evidence. The scale is nevertheless central to the story because it suggests the operation was not confined to a few low-traffic test sites.
That is the necessary summary. Rebuilding Futurism’s reporting paragraph by paragraph would only reproduce the work. Its investigation contains the site histories, interviews and evidence behind the finding.
Why distribution systems rewarded it
Acquiring an existing publication buys more than a domain name. It can bring years of backlinks, indexed pages, entity associations and audience signals. Those are imperfect proxies for trust, but search and recommendation systems use proxies because they cannot commission an editor to assess every new page.
High-volume production adds another advantage. A network that publishes across many topics can cover more queries, react to trends faster and generate more chances for one article to enter Search or Google Discover. Even if the average page performs poorly, the portfolio can produce large aggregate traffic.
The system is therefore rewarding observable signals: topical coverage, freshness, domain history, engagement and demand matching. It may not reliably observe whether the named author exists, whether the reporting happened, or whether the publisher would correct an error.
Google’s guidance on generative AI content does not ban AI-assisted production. It warns against using automation primarily to manipulate rankings and points publishers back to accuracy, quality and relevance. The weakness is enforcement at scale. A policy can describe useful content more easily than an algorithm can distinguish it from well-formatted imitation.
Discover adds another layer because recommendation traffic can be volatile and large. A publisher does not need a reader to search for its brand. It needs a page that matches a predicted interest at the right moment. That favors breadth and speed, the same qualities automation makes cheap.
The line that actually matters
AI use is not the decisive line. Deception and accountability are.
A legitimate AI-assisted publisher can name the humans responsible for a story, show where claims came from, distinguish reporting from synthesis, correct mistakes and explain how automation is used. The tools may help research, transcription, drafting, editing, translation or production. Accountability remains attached to real people and a real organization.
A fabricated byline breaks that chain. It tells the reader that a person with an identity and professional history stands behind the work when no such person exists. If the article also implies interviews, expertise or observation that never occurred, the problem expands from undisclosed automation to fabricated reporting.
Verifiable sourcing is the second line. Links alone are not enough, but they let readers test important claims. Original reporting should preserve notes, recordings or documents. Analysis should identify which facts come from others and which conclusions belong to the publisher.
Correction practices are the third line. A credible publisher needs a visible way to report errors, a process for reviewing them and a record of material changes. A network optimized only for output has little incentive to revisit yesterday’s pages after today’s traffic opportunity arrives.
Finally, an accountable publisher should disclose material AI involvement in language readers can understand. A generic footer does not excuse false sourcing or invented authors, but it establishes a baseline of honesty about the production process.
ContentGrip’s own position
ContentGrip uses AI agents in its editorial workflow. Every article is reviewed by an editor, published under a named byline and expected to link the sources behind factual claims. AI assistance does not replace the editor’s responsibility for what appears on the site.
That standard is not proof that every decision will be correct. It is a commitment that a real person and publication remain answerable when something is not.
What content teams should do
Start with acquired domains. Before buying a publication for its traffic or authority, audit its archive, ownership changes, backlink profile, bylines and correction history. A domain can carry useful equity and hidden liabilities at the same time.
Require named accountability for every page. That does not mean one writer must type every sentence. It means readers and colleagues can identify who approved the article and who will respond if it is wrong.
Publish an AI disclosure policy that describes actual practice. State where AI can be used, which tasks require human review, what cannot be automated and how the organization handles generated images, quotes, bylines and corrections.
Separate production targets from editorial evidence. A quota can reward speed even when a style guide asks for accuracy. Track the share of articles with primary sources, original interviews, documented review and substantive corrections, not only output and pageviews.
Content teams should also prepare for distribution volatility. The same systems that reward a scaled network can reduce its reach after a policy or ranking update. A volume-only operation with no loyal audience, recognizable writers or direct relationship has little to fall back on.
ContentGrip’s earlier report on AI slop as a brand-safety problem examined what happens when low-quality synthetic pages become ad inventory. The Brown Brothers Media investigation exposes the publishing side of the same market: cheap supply can look successful when distribution systems reward the signals around the content more consistently than the accountability behind it.
The uncomfortable lesson is that legitimate publishers cannot define themselves merely by saying they use humans or avoid AI. They need visible standards that readers, advertisers and platforms can test. The difference is not the presence of automation. It is whether the publication is honest about who made the work, how the claims were verified and who answers for the result.
This article is produced by ContentGrow. We’re building branded media outlets for B2B companies. Interested in learning more? Learn more.